v8.21 testing & results

Message boards : SETI@home Enhanced : v8.21 testing & results
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60308 - Posted: 3 Dec 2016, 21:20:37 UTC

I've downloaded r3578 and re-activated s@h Beta.  BOINC downloaded setiathome_8.21_windows_intelx86__opencl_nvidia_SoG.exe.  Should I be testing this or MB8_win_x86_SSE3_OpenCL_NV_SoG_r3578.exe?  Or are they identical (file sizes are =)?
ID: 60308 · Report as offensive
HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Dec 09
Posts: 74
Credit: 1,248,766
RAC: 0
United States
Message 60309 - Posted: 3 Dec 2016, 22:16:27 UTC - in response to Message 60308.  
Last modified: 3 Dec 2016, 22:20:20 UTC

I've downloaded r3578 and re-activated s@h Beta.  BOINC downloaded setiathome_8.21_windows_intelx86__opencl_nvidia_SoG.exe.  Should I be testing this or MB8_win_x86_SSE3_OpenCL_NV_SoG_r3578.exe?  Or are they identical (file sizes are =)?

For Windows builds
Raistmer said : I expect r3571 for non-SoG and r3578 for SoG builds, please docheck that.
For Linux builds
Urs said: r3566 for iGPU and r3570 for all others.

The build number should be present in the stderr. You can view it from your slow folder once the app start running if you don't want to want for a task to finish.

Unless explicitly directed to do otherwise by one of the developers you should always be using the version provided by the server for Beta.
ID: 60309 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60310 - Posted: 3 Dec 2016, 22:34:47 UTC - in response to Message 60309.  

I should have waited for the Stderr output.

Unless explicitly directed to do otherwise by one of the developers you should always be using the version provided by the server for Beta.
Understood


On a side note, there is something wrong with this site since the 'change'. When I opened this thread, I saw your post as:

For Windows builds
Raistmer said : I expect r3571 for non-SoG and r3578 for SoG builds, please docheck that.
For Linux builds
Urs said: r3566 for iGPU and r3570 for all others.

Unless explicitly directed to do otherwise by one of the developers you should always be using the version provided by the server for Beta.


But when I selected "Reply", your post looked like this:

For Windows builds
Raistmer said : I expect r3571 for non-SoG and r3578 for SoG builds, please docheck that.
For Linux builds
Urs said: r3566 for iGPU and r3570 for all others.

The build number should be present in the stderr. You can view it from your slow folder once the app start running if you don't want to want for a task to finish.

Unless explicitly directed to do otherwise by one of the developers you should always be using the version provided by the server for Beta.
ID: 60310 · Report as offensive
HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Dec 09
Posts: 74
Credit: 1,248,766
RAC: 0
United States
Message 60311 - Posted: 4 Dec 2016, 0:49:06 UTC - in response to Message 60310.  
Last modified: 4 Dec 2016, 0:49:31 UTC

I wonder if there is an issue with edited posts displaying the most resent information, or if there was an internet caching issue.

It may be hard to reproduce that kind of issue.
ID: 60311 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60318 - Posted: 5 Dec 2016, 1:12:42 UTC

The server only provided me with VLARs.

GPU is not OC’d, it’s strictly stock with default vendor settings.  1 WU/GPU.

Name:                     GeForce GTX 950
Vendor:                   NVIDIA Corporation
Driver version:           359.00
Version:                  OpenCL 1.2 CUDA

>  stock
    Task 25475727
    WU true angle range : 0.007806
    Run time : 15m 44s
    CPU time : 15m 34s

    Task 25475723
    WU true angle range : 0.007806
    Run time : 15m 10s
    CPU time : 14m 59s

    Task 25475668
    WU true angle range : 0.007806
    Run time : 15m 03s
    CPU time : 14m 52s

    Task 25483471
    WU true angle range : 0.009111
    Run time : 15m 07s
    CPU time : 14m 44s

    NOTES -- GUI lag unacceptable while playing a non-action game.  Also, noticeable while surfing web and switching between apps.


>  -tt 30

    Task 25483559
    WU true angle range : 0.009111
    Run time : 15m 06s
    CPU time : 14m 53s

    Task 25483337
    WU true angle range : 0.009111
    Run time : 15m 08s
    CPU time : 14m 59s

    Task 25483246
    WU true angle range : 0.009111
    Run time : 15m 11s
    CPU time : 14m 59s

    NOTES -- Mouse movement is choppy.  Toggling between screens lags.


>  -tt 60

    Task 25483602
    WU true angle range : 0.009111
    Run time : 15m 05s
    CPU time : 14m 51s

    Task 25483693
    WU true angle range : 0.009111
    Run time : 15m 02s
    CPU time : 14m 53s

    Task 25483154
    WU true angle range : 0.009111
    Run time : 15m 02s
    CPU time : 14m 53s

    NOTES -- Mouse movement is choppy.  Toggling between screens lags.


>  -tt 90

    Task 25482672
    WU true angle range : 0.007806
    Run time : 15m 03s
    CPU time : 14m 46s

    Task 25483553
    WU true angle range : 0.009111
    Run time : 15m 08s
    CPU time : 14m 50s

    Task 25483555
    WU true angle range : 0.009111
    Run time : 15m 06s
    CPU time : 14m 49s

    Task 25483718
    WU true angle range : 0.009111
    Run time : 15m 12s
    CPU time : 15m 02s

    NOTES -- Could not play "Hover."  Win7 notified that game stopped working, then locked up rig.


>  -tt 500

    Task 25483173
    WU true angle range : 0.009111
    Run time : 15m 03s
    CPU time : 14m 55s

    Task 25483160
    WU true angle range : 0.009111
    Run time : 15m 01s
    CPU time : 14m 53s

    Task 25483688
    WU true angle range : 0.009111
    Run time : 15m 02s
    CPU time : 14m 53s

    Task 25482948
    WU true angle range : 0.007806
    Run time : 14m 59s
    CPU time : 14m 50s

    Task 25483673
    WU true angle range : 0.009111
    Run time : 15m 00s
    CPU time : 14m 51s

    Task 25483575
    WU true angle range : 0.009111
    Run time : 15m 02s
    CPU time : 14m 52

    NOTES -- There was some lag while surfing web and toggling among browser tabs.


>  -tt 1000

    Task 25483424
    WU true angle range : 0.009111
    Run time : 15m 06s
    CPU time : 14m 50s

    Task 25483664
    WU true angle range : 0.009111
    Run time : 15m 03s
    CPU time : 14m 44s

    Task 25483584
    WU true angle range : 0.009111
    Run time : 15m 03s
    CPU time : 14m 50s

    NOTES -- Still a GUI lag.  For example, it is noticeable with the animation of minimizing a window.


>  -sbs 256

    Task 25483740
    WU true angle range : 0.007977
    Run time : 15m 02s
    CPU time : 14m 50s

    Task 25483739
    WU true angle range : 0.007977
    Run time : 15m 02s
    CPU time : 14m 51s

    NOTES -- Still a GUI lag.  GUI lag noticeable while playing a non-action game.


I would also like to point that on Main, I'm running v8.20 (opencl_nvidia_SoG) with the following cmdline:

-sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4 -high_prec_timer -high_perf
and the VLAR run times are from ~1232 sec to ~1384 sec.  At this point, VLARs on Beta are running from ~899 sec to ~911 sec (nearly identical ARs).

For quite some hours, I've been getting this message:

12/4/2016 8:09:13 PM | SETI@home Beta Test | Sending scheduler request: To fetch work.
12/4/2016 8:09:13 PM | SETI@home Beta Test | Requesting new tasks for NVIDIA GPU
12/4/2016 8:09:15 PM | SETI@home Beta Test | Scheduler request completed: got 0 new tasks
ID: 60318 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60319 - Posted: 5 Dec 2016, 15:12:07 UTC - in response to Message 60318.  

To reduce GUI lag you need decrease target time value (-tt N ).
-tt 1000 means 1s of continuous kernel run allowed - lag just guarantied for such setting if enough data for GPU (not too fast GPU) can be provided at all.

Try -tt 15 for example. Then adjust -period_iterations_num N (here N should be increased to reduce lag)
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60319 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60332 - Posted: 8 Dec 2016, 1:51:16 UTC - in response to Message 60319.  

GPU is not OC’d, it’s strictly stock with default vendor settings.  1 WU/GPU.

Name:                     GeForce GTX 950
Vendor:                   NVIDIA Corporation
Driver version:           359.00
Version:                  OpenCL 1.2 CUDA

>  -tt 15 -period_iterations_num 20
    Task 25492724
    WU true angle range : 0.007806
    Run time : 15m 19s
    CPU time : 15m 07s

    Task 25492620
    WU true angle range : 0.007806
    Run time : 15m 17s
    CPU time : 15m 08s

    Task 25492397
    WU true angle range : 0.007806
    Run time : 15m 20s
    CPU time : 15m 08s

    Task 25492581
    WU true angle range : 0.007806
    Run time : 15m 20s
    CPU time : 15m 07s

    Task 25492869
    WU true angle range : 0.009111
    Run time : 15m 21s
    CPU time : 15m 08s

    NOTES -- Slightly noticeable mouse lag while playing a non-action game.


>  -tt 15 -period_iterations_num 30

    Task 25492955
    WU true angle range : 0.009111
    Run time : 15m 22s
    CPU time : 15m 07s

    Task 25492905
    WU true angle range : 0.009111
    Run time : 15m 23s
    CPU time : 15m 07s

    Task 25493283
    WU true angle range : 0.009111
    Run time : 15m 23s
    CPU time : 15m 07s

    [url=]Task [/url]
    WU true angle range :
    Run time :
    CPU time :

    [url=]Task [/url]
    WU true angle range :
    Run time :
    CPU time :

    NOTES -- Slightly noticeable mouse lag while playing a non-action game.


>  -tt 15 -period_iterations_num 40

    Task 25493308
    WU true angle range : 0.009111
    Run time : 15m 22s
    CPU time : 15m 08s

    Task 25493526
    WU true angle range : 0.009111
    Run time : 15m 28s
    CPU time : 15m 05s

    Task 25493930
    WU true angle range : 0.007977
    Run time : 15m 18s
    CPU time : 15m 08s

    Task 25493898
    WU true angle range : 0.007977
    Run time : 15m 20s
    CPU time : 15m 09s

    Task 25494160
    WU true angle range : 0.007977
    Run time : 15m 19s
    CPU time : 15m 08s

    NOTES -- Improved response from mouse and scroll wheel


>  -tt 15 -period_iterations_num 50

    Task 25493955
    WU true angle range : 0.007977
    Run time : 15m 22s
    CPU time : 15m 08s

    Task 25494415
    WU true angle range : 0.007977
    Run time : 15m 20s
    CPU time : 15m 08s

    Task 25494857
    WU true angle range : 0.008056
    Run time : 15m 22s
    CPU time : 15m 10s

    Task 25494987
    WU true angle range : 0.008056
    Run time : 15m 20s
    CPU time : 15m 09s

    NOTES -- No noticeable lags


The most notable aspect of v8.21 is that it is completing WUs in 25% - 36% less time for nearly identical ARs than what I'm running on Main.

@Raistmer
I'm going to test some other cmdline options.  If you have some specific options that you'd like me to test, please post them.

ID: 60332 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60334 - Posted: 8 Dec 2016, 7:50:52 UTC - in response to Message 60332.  

Please look for memory usage with different -tt N values and -sbs N values.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60334 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60339 - Posted: 9 Dec 2016, 1:34:43 UTC - in response to Message 60334.  
Last modified: 9 Dec 2016, 1:34:55 UTC

Please look for memory usage with different -tt N values and -sbs N values.
Is the memory usage located on the Task page above the Stderr output, or within the Stderr output?
ID: 60339 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60343 - Posted: 9 Dec 2016, 6:29:10 UTC - in response to Message 60339.  
Last modified: 9 Dec 2016, 6:29:22 UTC

Please look for memory usage with different -tt N values and -sbs N values.
Is the memory usage located on the Task page above the Stderr output, or within the Stderr output?

Actual fields in bold:

http://setiathome.berkeley.edu/result.php?resultid=5342593694
Имя 20au09ac.29508.476.15.42.66_0
Задача 2352756579
Создан 7 Dec 2016, 10:37:37 UTC
Отправлен 7 Dec 2016, 15:14:13 UTC
Крайний срок отчёта 28 Jan 2017, 23:41:57 UTC
Получен 9 Dec 2016, 2:43:58 UTC
Состояние сервера Завершено
Результат выполнения Успех
Состояние клиента Готово
Статус выхода 0 (0x00000000)
ID компьютера 4517643
Время выполнения 2 часов 50 мин. 19 сек.
Время ЦП 2 часов 39 мин. 55 сек.
Состояние проверки Правильный
Очки 97.14
Пиковая производительность устройства, FLOPS 0.00 GFLOPS
Версия приложения SETI@home v8
Анонимная платформа (Тип ЦП)
Peak working set size 38.99 MB
Peak swap size 47.98 MB

Peak disk usage 0.03 MB

News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60343 · Report as offensive
HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Dec 09
Posts: 74
Credit: 1,248,766
RAC: 0
United States
Message 60345 - Posted: 9 Dec 2016, 20:57:42 UTC - in response to Message 60343.  

Please look for memory usage with different -tt N values and -sbs N values.
Is the memory usage located on the Task page above the Stderr output, or within the Stderr output?

Actual fields in bold:

http://setiathome.berkeley.edu/result.php?resultid=5342593694
Имя 20au09ac.29508.476.15.42.66_0
Задача 2352756579
Создан 7 Dec 2016, 10:37:37 UTC
Отправлен 7 Dec 2016, 15:14:13 UTC
Крайний срок отчёта 28 Jan 2017, 23:41:57 UTC
Получен 9 Dec 2016, 2:43:58 UTC
Состояние сервера Завершено
Результат выполнения Успех
Состояние клиента Готово
Статус выхода 0 (0x00000000)
ID компьютера 4517643
Время выполнения 2 часов 50 мин. 19 сек.
Время ЦП 2 часов 39 мин. 55 сек.
Состояние проверки Правильный
Очки 97.14
Пиковая производительность устройства, FLOPS 0.00 GFLOPS
Версия приложения SETI@home v8
Анонимная платформа (Тип ЦП)
Peak working set size 38.99 MB
Peak swap size 47.98 MB

Peak disk usage 0.03 MB

So we need to be looking at application memory & not GPU memory usage?
I was reporting GPU memory values from GPUz previously.
ID: 60345 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60346 - Posted: 9 Dec 2016, 22:34:43 UTC - in response to Message 60345.  

So we need to be looking at application memory & not GPU memory usage?
I was reporting GPU memory values from GPUz previously.

Ah, I see...
Yes, host/system RAM, not GPU.

GPU will allocate more with -sbs N increase, it's definitely correct behavior.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60346 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60353 - Posted: 11 Dec 2016, 16:29:31 UTC - in response to Message 60343.  

Name:                     GeForce GTX 950
Vendor:                   NVIDIA Corporation
Driver version:           359.00
Version:                  OpenCL 1.2 CUDA

>  -tt 30 -sbs 192
    Task 25495210
    WU true angle range : 0.008056
    Peak working set size : 143.75 MB
    Peak swap size : 171.43 MB
    Run time : 15m 08s
    CPU time : 14m 54s

    Task 25495128
    WU true angle range : 0.008056
    Peak working set size : 143.28 MB
    Peak swap size : 170.91 MB
    Run time : 15m 10s
    CPU time : 14m 56s

    Task 25495048
    WU true angle range : 0.008056
    Peak working set size : 143.70 MB
    Peak swap size : 171.40 MB
    Run time : 15m 09s
    CPU time : 14m 56s


>  -tt 90 -sbs 192

    Task 25495532
    WU true angle range : 0.008056
    Peak working set size :  141.93 MB
    Peak swap size : 169.64 MB
    Run time : 15m 03s
    CPU time : 14m 58s

    Task 25495647
    WU true angle range : 0.008056
    Peak working set size : 141.93 MB
    Peak swap size : 169.61 MB
    Run time : 15m 00s
    CPU time : 14m 55s

    Task 25495762
    WU true angle range : 0.25495762
    Peak working set size : 141.98 MB
    Peak swap size : 169.63 MB
    Run time : 15m 01s
    CPU time : 14m 56s


>  -tt 180 -sbs 192

    Task 25495776
    WU true angle range : 0.008056
    Peak working set size : 141.84 MB
    Peak swap size : 169.55 MB
    Run time : 15m 00s
    CPU time : 14m 54s

    Task 25495971
    WU true angle range : 0.008056
    Peak working set size : 132.00 MB
    Peak swap size : 159.67 MB
    Run time : 15m 10s
    CPU time : 15m 02s

    Task 25495925
    WU true angle range : 0.008056
    Peak working set size : 142.44 MB
    Peak swap size : 170.13 MB
    Run time : 14m 59s
    CPU time : 14m 55s

    Task 25495966
    WU true angle range : 0.008056
    Peak working set size : 142.00 MB
    Peak swap size : 169.61 MB
    Run time : 14m 57s
    CPU time : 14m 53s


>  -tt 500 -sbs 192

    Task 25496334
    WU true angle range : 0.009111
    Peak working set size : 138.09 MB
    Peak swap size : 165.79 MB
    Run time : 15m 11s
    CPU time : 15m 03s

    Task 25495956
    WU true angle range : 0.008056
    Peak working set size : 141.93 MB
    Peak swap size : 169.65 MB
    Run time : 14m 58s
    CPU time : 14m 53s

    Task 25495826
    WU true angle range : 0.008056
    Peak working set size : 141.98 MB
    Peak swap size : 169.70 MB
    Run time : 14m 57s
    CPU time : 14m 52s

More to follow.

ID: 60353 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60354 - Posted: 11 Dec 2016, 17:14:57 UTC - in response to Message 60343.  

>  -tt 1000 -sbs 192
    Task 25496550
    WU true angle range : 009111
    Peak working set size : 142.16 MB
    Peak swap size : 169.80 MB
    Run time : 14m 56s
    CPU time : 14m 53s

    Task 25496409
    WU true angle range : 0.009111
    Peak working set size : 142.27 MB
    Peak swap size : 169.96 MB
    Run time : 15m 03s
    CPU time : 14m 56s

    Task 25496672
    WU true angle range : 0.009111
    Peak working set size : 141.95 MB
    Peak swap size : 169.65 MB
    Run time : 15m 04s
    CPU time : 14m 56s

    Task 25496569
    WU true angle range : 0.009111
    Peak working set size : 142.32 MB
    Peak swap size : 169.96 MB
    Run time : 15m 01s
    CPU time : 14m 55s

ID: 60354 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60360 - Posted: 12 Dec 2016, 8:03:16 UTC - in response to Message 60354.  
Last modified: 12 Dec 2016, 8:06:49 UTC

Target kernel sequence time set to 1000ms

Fftlength=32,pass=3:Tune: sum=30500.9(ms); min=5.767(ms); max=87.06(ms); mean=42.78(ms); s_mean=66.61; sleep=60(ms); delta=822; N=713; usual
Fftlength=32,pass=4:Tune: sum=23455.3(ms); min=5.053(ms); max=71.9(ms); mean=34.34(ms); s_mean=49.47; sleep=45(ms); delta=781; N=683; usual
Fftlength=32,pass=5:Tune: sum=20563.2(ms); min=5.376(ms); max=65.47(ms); mean=31.98(ms); s_mean=54.56; sleep=45(ms); delta=738; N=643; usual
Fftlength=64,pass=3:Tune: sum=28971(ms); min=2.777(ms); max=84.57(ms); mean=37.05(ms); s_mean=54.52; sleep=45(ms); delta=836; N=782; usual
Fftlength=64,pass=4:Tune: sum=22277.6(ms); min=2.627(ms); max= 69(ms); mean=29.74(ms); s_mean=40.69; sleep=30(ms); delta=798; N=749; usual
Fftlength=64,pass=5:Tune: sum=19974.6(ms); min=2.704(ms); max=65.62(ms); mean=28.29(ms); s_mean=58.3; sleep=60(ms); delta=753; N=706; usual
Fftlength=128,pass=3:Tune: sum=28898.1(ms); min=1.496(ms); max=83.39(ms); mean=34.69(ms); s_mean=59.24; sleep=60(ms); delta=860; N=833; usual
Fftlength=128,pass=4:Tune: sum=22333(ms); min=1.326(ms); max=68.78(ms); mean=28.02(ms); s_mean=45.59; sleep=45(ms); delta=821; N=797; usual
Fftlength=128,pass=5:Tune: sum=20065.2(ms); min=1.397(ms); max=67.32(ms); mean=26.51(ms); s_mean=40.76; sleep=30(ms); delta=780; N=757; usual
Fftlength=256,pass=3:Tune: sum=50415.6(ms); min=0.6968(ms); max=162.6(ms); mean=56.39(ms); s_mean=156.2; sleep=150(ms); delta=907; N=894; high_perf
Fftlength=256,pass=4:Tune: sum=11678.4(ms); min=0.6703(ms); max=51.43(ms); mean=18.69(ms); s_mean=50.22; sleep=45(ms); delta=863; N=625; usual
Fftlength=256,pass=5:Tune: sum=10512.9(ms); min=0.7098(ms); max=47.37(ms); mean=18.09(ms); s_mean=46.24; sleep=45(ms); delta=818; N=581; usual
Fftlength=512,pass=3:Tune: sum=65662.5(ms); min=0.3574(ms); max=79.09(ms); mean=57.2(ms); s_mean=76.26; sleep=75(ms); delta=1154; N=1148; high_perf
Fftlength=512,pass=4:Tune: sum=2874.24(ms); min=0.3548(ms); max=25.51(ms); mean=9.212(ms); s_mean=24.33; sleep=15(ms); delta=1132; N=312; usual
Fftlength=512,pass=5:Tune: sum=2458.35(ms); min=0.3927(ms); max=20.43(ms); mean=8.477(ms); s_mean=20.25; sleep=15(ms); delta=1109; N=290; usual
Fftlength=1024,pass=3:Tune: sum=69088.7(ms); min=0.2148(ms); max=38.74(ms); mean=34.98(ms); s_mean=37.68; sleep=30(ms); delta=1978; N=1975; high_perf
Fftlength=1024,pass=4:Tune: sum=719.176(ms); min=0.1874(ms); max=11.98(ms); mean=4.61(ms); s_mean=10.33; sleep=0(ms); delta=1968; N=156; usual
Fftlength=1024,pass=5:Tune: sum=639.539(ms); min=0.2054(ms); max=10.55(ms); mean=4.38(ms); s_mean=10.33; sleep=0(ms); delta=1957; N=146; usual
Fftlength=2048,pass=3:Tune: sum=52003.1(ms); min=5.659(ms); max=14.9(ms); mean=13.91(ms); s_mean=13.97; sleep=15(ms); delta=1; N=3739; high_perf
Fftlength=4096,pass=3:Tune: sum=48331.8(ms); min=2.857(ms); max=6.956(ms); mean=6.464(ms); s_mean=6.47; sleep=0(ms); delta=1; N=7477; high_perf
Fftlength=8192,pass=3:Tune: sum=37624.3(ms); min=2.346(ms); max=3.292(ms); mean=2.516(ms); s_mean=2.54; sleep=0(ms); delta=1; N=14953; usual


Seems only few kernels on your GPU can exceed ~100ms length so any -tt N with N bigger than 150 - 200 will act the same (saturated).
(for details look: http://lunatics.kwsn.info/index.php/topic,1808.msg61251.html#msg61251 )

Try to lower -tt value instead.
Let say
-tt 10
-tt 5
-tt 1

I expect drop in performance for those values, but what will be with memory consumption? That's the aim of testing.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60360 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60365 - Posted: 13 Dec 2016, 1:15:45 UTC - in response to Message 60360.  

Seems only few kernels on your GPU can exceed ~100ms length so any -tt N with N bigger than 150 - 200 will act the same (saturated).
(for details look: http://lunatics.kwsn.info/index.php/topic,1808.msg61251.html#msg61251 )
I saw that post on Main, but I haven't read it yet.  I've read everything else on that page.

Try to lower -tt value instead.
Let say
-tt 10
-tt 5
-tt 1

I expect drop in performance for those values, but what will be with memory consumption? That's the aim of testing.

>  -tt 30 -sbs 256
    Task 25497002
    WU true angle range : 0.007977
    Peak working set size : 143.72 MB
    Peak swap size : 171.40 MB
    Run time : 15m 05s
    CPU time : 15m 01s

    Task 25497272
    WU true angle range : 0.007977
    Peak working set size : 143.70 MB
    Peak swap size : 171.39 MB
    Run time : 15m 09s
    CPU time : 15m 06s

    Task 25497714
    WU true angle range : 0.007806
    Peak working set size : 143.80 MB
    Peak swap size : 171.52 MB
    Run time : 15m 05s
    CPU time : 15m 02s


>  -tt 1 -sbs 256

    Task 25497717
    WU true angle range : 0.007806
    Peak working set size : 321.66 MB
    Peak swap size : 349.45 MB
    Run time : 16m 47s
    CPU time : 16m 43s

    Task 25497508
    WU true angle range : 0.007977
    Peak working set size : 322.57 MB
    Peak swap size : 350.38 MB
    Run time : 16m 52s
    CPU time : 16m 48s

    Task 25497753
    WU true angle range : 0.007806
    Peak working set size : 321.95 MB
    Peak swap size : 349.72 MB
    Run time : 16m 49s
    CPU time : 16m 46s


>  -tt 5 -sbs 256

    Task 25497510
    WU true angle range : 0.007977
    Peak working set size : 185.30 MB
    Peak swap size : 213.04 MB
    Run time : 15m 39s
    CPU time : 15m 35s

    Task 25496970
    WU true angle range : 0.007977
    Peak working set size : 184.92 MB
    Peak swap size : 212.81 MB
    Run time : 15m 43s
    CPU time : 15m 38s

    Task 25497704
    WU true angle range : 0.007806
    Peak working set size : 184.81 MB
    Peak swap size : 212.62 MB
    Run time : 15m 42s
    CPU time : 15m 36s


>  -tt 10 -sbs 256

    Task 25497672
    WU true angle range : 0.007806
    Peak working set size : 159.49 MB
    Peak swap size : 187.25 MB
    Run time : 15m 38s
    CPU time : 15m 31s

    Task 25497492
    WU true angle range : 0.007977
    Peak working set size : 159.26 MB
    Peak swap size : 187.06 MB
    Run time : 15m 32s
    CPU time : 15m 28s

    Task 25497890
    WU true angle range : 0.007806
    Peak working set size : 159.35 MB
    Peak swap size : 187.04 MB
    Run time : 15m 35s
    CPU time : 15m 28s


The server finally graced me with a different AR.  See the first and third WUs below, they were fast.

>  -tt 1 -sbs 192

    Task 25559432
    WU true angle range : 0.848491
    Peak working set size : 105.63 MB
    Peak swap size : 135.22 MB
    Run time : 08m 35s
    CPU time : 05m 03s

    Task 25558820
    WU true angle range : 0.010364
    Peak working set size : 433.75 MB
    Peak swap size : 461.62 MB
    Run time : 26m 25s
    CPU time : 26m 14s

    Task 25559398
    WU true angle range : 0.848491
    Peak working set size : 105.74 MB
    Peak swap size : 133.47 MB
    Run time : 08m 33s
    CPU time : 05m 02s

More to follow.

ID: 60365 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60367 - Posted: 13 Dec 2016, 14:58:36 UTC

>  -tt 5 -sbs 192

    Task 25559440
    WU true angle range : 0.848491
    Peak working set size : 1,009.49 MB
    Peak swap size : 1,072.76 MB
    Run time : 08m 09s
    CPU time : 04m 43s

    Task 25559356
    WU true angle range : 0.848491
    Peak working set size : 103.78 MB
    Peak swap size : 131.46 MB
    Run time : 08m 23s
    CPU time : 04m 53s

    Task 25559290
    WU true angle range : 0.848491
    Peak working set size : 103.82 MB
    Peak swap size : 131.53 MB
    Run time : 08m 32s
    CPU time : 05m 02s

    Task 25559281
    WU true angle range : 0.848491
    Peak working set size : 104.01 MB
    Peak swap size : 131.68 MB
    Run time : 08m 34s
    CPU time : 05m 04s

    NOTES -- Why such a huge difference in mem usage between first and other WUs?

ID: 60367 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 12 Jan 16
Posts: 38
Credit: 289,647
RAC: 0
United States
Message 60369 - Posted: 14 Dec 2016, 2:08:41 UTC

>  -tt 10 -sbs 192

    Task 25559257
    WU true angle range : 0.848491
    Peak working set size : 105.14 MB
    Peak swap size : 135.16 MB
    Run time : 08m 33s
    CPU time : 05m 02s

    Task 25559431
    WU true angle range : 0.848491
    Peak working set size : 103.78 MB
    Peak swap size : 131.45 MB
    Run time : 08m 36s
    CPU time : 05m 05s

    Task 25559320
    WU true angle range : 0.848491
    Peak working set size : 103.32 MB
    Peak swap size : 134.91 MB
    Run time : 08m 42s
    CPU time : 05m 11s


>  -high_perf

    Task 25559386
    WU true angle range : 0.848491
    Peak working set size : 103.19 MB
    Peak swap size : 134.89 MB
    Run time : 08m 13s
    CPU time : 04m 44s

    Task 25559407
    WU true angle range : 0.848491
    Peak working set size : 102.82 MB
    Peak swap size : 130.53 MB
    Run time : 08m 31s
    CPU time : 04m 44s

    Task 25559168
    WU true angle range : 0.848491
    Peak working set size : 102.84 MB
    Peak swap size : 130.57 MB
    Run time : 08m 31s
    CPU time : 04m 43s


>  -sbs 192 -high_perf

    Task 25494739
    WU true angle range : 0.008056
    Peak working set size : 141.54 MB
    Peak swap size : 169.27 MB
    Run time : 15m 01s
    CPU time : 14m 50s

    Task 25494892
    WU true angle range : 0.008056
    Peak working set size : 141.09 MB
    Peak swap size : 168.86 MB
    Run time : 15m 00s
    CPU time : 14m 49s

    Task 25494718
    WU true angle range : 0.008056
    Peak working set size : 141.50 MB
    Peak swap size : 169.29 MB
    Run time : 15m 04s
    CPU time : 14m 47s


>  -tt 1 -sbs 192 -high_perf

    Task 25559389
    WU true angle range : 0.848491
    Peak working set size : 104.55 MB
    Peak swap size : 104.55 MB
    Run time : 08m 33s
    CPU time : 04m 47s

    Task 25559392
    WU true angle range : 0.848491
    Peak working set size : 104.53 MB
    Peak swap size : 132.27 MB
    Run time : 08m 35s
    CPU time : 04m 53s

    Task 25559408
    WU true angle range : 0.848491
    Peak working set size : 106.63 MB
    Peak swap size : 106.63 MB
    Run time : 08m 14s
    CPU time : 04m 29s


>  -tt 5 -sbs 192 -high_perf

    Task 25559315
    WU true angle range : 0.848491
    Peak working set size : 104.30 MB
    Peak swap size : 132.03 MB
    Run time : 08m 08s
    CPU time : 04m 25s

    Task 25559178
    WU true angle range : 0.848491
    Peak working set size : 103.74 MB
    Peak swap size : 135.08 MB
    Run time : 08m 29s
    CPU time : 04m 50s

    Task 25559268
    WU true angle range : 0.848491
    Peak working set size : 104.34 MB
    Peak swap size : 132.03 MB
    Run time : 08m 32s
    CPU time : 04m 44s


>  -tt 10 -sbs 192 -high_perf

    Task 25559400
    WU true angle range : 0.848491
    Peak working set size : 103.37 MB
    Peak swap size : 131.10 MB
    Run time : 08m 32s
    CPU time : 05m 01s

    Task 25559400
    WU true angle range : 0.848491
    Peak working set size : 103.37 MB
    Peak swap size : 131.10 MB
    Run time : 08m 32s
    CPU time : 05m 01s

    Task 25559327
    WU true angle range : 0.848491
    Peak working set size : 107.64 MB
    Peak swap size : 134.91 MB
    Run time : 08m 33s
    CPU time : 05m 01s

ID: 60369 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 60370 - Posted: 14 Dec 2016, 8:07:57 UTC - in response to Message 60367.  

>  -tt 5 -sbs 192
NOTES -- Why such a huge difference in mem usage between first and other WUs?[/list]

Single thing I could spot in stderr- task was restarted from checkpoint.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 60370 · Report as offensive
tullio
Volunteer tester

Send message
Joined: 9 May 16
Posts: 11
Credit: 431,387
RAC: 0
Italy
Message 60372 - Posted: 14 Dec 2016, 17:05:46 UTC
Last modified: 14 Dec 2016, 17:06:05 UTC

I am crunching v8.22 on my Linux box.
Tullio
ID: 60372 · Report as offensive
1 · 2 · Next

Message boards : SETI@home Enhanced : v8.21 testing & results


 
©2021 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.